2009

Sat, Dec 19

<00:42 EEST> As the academic term is practically finished — lectures are over but there is one essay left to write — I continue keeping up with Linux graphics development. I have mentioned Gallium3D before, and it has now arrived in Gentoo with Mesa version 7.7_rc2. It is still behind a USE flag due to the highly experimental nature. In fact, only 7.7_rc3 managed to compile on my systems when using Gallium, but fortunately that version works fine on all of my three machines that are set up for OpenGL. That makes two Intels and a Radeon. In fact, nunu and sable also have some level of OpenGL hardware, but not in practical use.

However, the Gallium-enabled Mesa runs considerably slower than the old implementation, about half of the frame rate. This is not surprising as the technology is still being developed, but for now I have reverted back to the faster version, as I actually have professional uses for 3D acceleration. My chemistry studies have led me into molecular modelling and visualization, and so far the main software I have used on my own machines is Avogadro. In addition to basic modelling and optimization, it handles molecular mechanics with a few different force fields available.

Tue, Dec 15

<17:16 EEST> With the Linux kernel point updates 2.6.32.1 and 2.6.31.8 out, my laptop is still without the 2.6.32 series, as the modesetting bug has not bee fixed. So far I have not found any directly relevant bug reports, but Gentoo Wiki has pointed me to a rather comprehensive status report on Intel graphics. It seems that a similar blank-screen problem happened before when KMS was introduced in 2.6.29, and there are further mentions of 8xx-specific bugs. The page also mentions "UMS broken on 855 with xserver 1.7" which was the original reason I switched to KMS. So there are no easy fixed yet, but at least there are comprehensive tests being done with the 855, and I am not alone here. Of course, I should do some bug reporting myself, but I do not know where to start, and at this time of the academic year I have other concerns. I think I can just about do some revision and TeXnical writing, even without the latest kernel :)

<17:52 EEST> Well, now that I got into Freedesktop Bugzilla, I found just the issue, complete with a patch. It looks like a quick n' dirty fix that simply overrides the lid status detection, so it is understandable that the vanilla kernel will end up with something more comprehensive. But for now, I can continue my studies in peace, with the blazing new shine of 2.6.32.1 :-j

Sat, Dec 5

<19:44 EEST> It's been a long and quiet November as far as What's Up is concerned, and there is still not much newsworthy action going on here, but I thought I'd share some recent Linux issues. Mostly about kernel modesetting, which I have been using succesfully on both of my active x86 Intel machines for a few weeks now. I had tried it some time in the past, and the last time I took the big dip it worked like charm.

There are a few things to consider if you want KMS. I have only tried it on Intel grahics, which is probably the stablest platform to try it on (and anything Linux graphics related :) though ATI seems to be coming along as well.

The hard bit, in a way, is that you cannot use anything like vesafb when a KMS DRM is loaded. The DRM has its own framebuffer though. I have found it easiest to build the DRM into the kernel and enable KMS by default. In case it does not work, just reboot into a safe kernel. Xorg should start with no special configuration, and you should find that there is no flickery lag when you switch back into a framebuffer console.

The reason I remembered to write now is that Linux 2.6.32 was released this Thursday, so I have been playing with the new kernel for a while. Unfortunately, the KMS DRM does not find any display modes on Sigmatrix with this kernel. On Hoo it works fine, which is expected as more recent Intel graphics seem to be better supported overall. Curiously enough, this release is supposed to fix something related:

8xx works again, since the regression with GEM's introduction back in .27

as reported by Kernel Newbies. In my case the 855GME has worked in the past with 2.6.31.6, but the latest kernel boots into a black screen — perhaps in an imitation of the black screen of death, recently seen in Windows :-/

<23:04 EEST> Recalling further on why I tried KMS a few weeks ago, it was because an updated Xorg then locked up the laptop using a non-KMS kernel (as reported by others). It became obvious just now as I tried a non-KMS 2.6.32. So I am back to 2.6.31.6 with modesetting.

Of course, KMS is not just about a fast switching between text console and X. Frankly, I have not followed all the details of this development, but it is likely that the non-KMS way will gradually fade away from the developer interests. Yet one of those things that many people seem to hate about Linux, in that the lack of standard interfaces hinders commercial and consumer uptake. On the flip side, Windows has all these problems because of its goals of remaining backwards compatible. Although when you have access to the source, you can run ancient software on a modern OS anyway :)

---

Risto A. Paju